[minicoredumper] minicoredumper coredump rotation

[John Ogness]

On 2019-11-04, Martyn Welch <martyn.welch at collabora.com> wrote:
>> On 2019-11-01, Martyn Welch <martyn.welch at collabora.com> wrote:
>>> We're following the patch series "Fix some autotools related issues
>>> and add coredump rotation" from March this year with interest,
>> 
>> Yes, the series by Mateusz looks good and includes mostly interesting
>> features. I have not taken the time to give them the serious audit
>> that is necessary for me to commit them.
>> 
>
> So reading between the lines a little here, is the plan to merge that
> series (assuming your audit doesn't throw up anything problematic)

Yes. The only patch that I already know that I do not want to take is:

    [PATCH 03/14] minicoredumper: read stack pointer from registers

The problem was fixed in the kernel (and backported to stable) so I see
no reason for a userspace workaround.

> and work on a pre-/post-script alternative to replace it at some
> point?

It will at least be seriously considered.

Keep in mind that it is already possible to do pre-/post-scripting by
having a script like this:

diff --git a/mcdwrap b/mcdwrap
new file mode 100755
index 0000000..7351624
--- /dev/null
+++ b/mcdwrap
@@ -0,0 +1,4 @@
+#!/bin/sh
+echo 'This is done before minicoredumper!'
+/usr/sbin/minicoredumper "$@"
+echo 'This is done after minicoredumper!'

Assuming the script (mcdwrap) is in /usr/sbin:

$ echo '|/usr/sbin/mcdwrap %P %u %g %s %t %h %e' | \
          sudo tee /proc/sys/kernel/core_pattern

If we really need the pre-/post-script feature built-in to
minicoredumper, it should be at _least_ as flexible as doing it with the
wrapper. Otherwise I don't see the point.

John Ogness